Asymptotic behavior of \ell_p-based Laplacian regularization in semi-supervised learning

Ahmed El Alaoui, Xiang Cheng, Aaditya Ramdas, Martin J. Wainwright, Michael I. Jordan
29th Annual Conference on Learning Theory, PMLR 49:879-906, 2016.

Abstract

Given a weighted graph with N vertices, consider a real-valued regression problem in a semi-supervised setting, where one observes n labeled vertices, and the task is to label the remaining ones. We present a theoretical study of \ell_p-based Laplacian regularization under a d-dimensional geometric random graph model. We provide a variational characterization of the performance of this regularized learner as N grows to infinity while n stays constant; the associated optimality conditions lead to a partial differential equation that must be satisfied by the associated function estimate \widehatf. From this formulation we derive several predictions on the limiting behavior the function \fhat, including (a) a phase transition in its smoothness at the threshold p = d + 1; and (b) a tradeoff between smoothness and sensitivity to the underlying unlabeled data distribution P. Thus, over the range p ≤d, the function estimate \widehatf is degenerate and “spiky,” whereas for p≥d+1, the function estimate \fhat is smooth. We show that the effect of the underlying density vanishes monotonically with p, such that in the limit p = ∞, corresponding to the so-called Absolutely Minimal Lipschitz Extension, the estimate \widehatf is independent of the distribution P. Under the assumption of semi-supervised smoothness, ignoring P can lead to poor statistical performance; in particular, we construct a specific example for d=1 to demonstrate that p=2 has lower risk than p=∞due to the former penalty adapting to P and the latter ignoring it. We also provide simulations that verify the accuracy of our predictions for finite sample sizes. Together, these properties show that p = d+1 is an optimal choice, yielding a function estimate \fhat that is both smooth and non-degenerate, while remaining maximally sensitive to P.

Cite this Paper


BibTeX
@InProceedings{pmlr-v49-elalaoui16, title = {Asymptotic behavior of $\ell_p$-based {L}aplacian regularization in semi-supervised learning}, author = {El Alaoui, Ahmed and Cheng, Xiang and Ramdas, Aaditya and Wainwright, Martin J. and Jordan, Michael I.}, booktitle = {29th Annual Conference on Learning Theory}, pages = {879--906}, year = {2016}, editor = {Feldman, Vitaly and Rakhlin, Alexander and Shamir, Ohad}, volume = {49}, series = {Proceedings of Machine Learning Research}, address = {Columbia University, New York, New York, USA}, month = {23--26 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v49/elalaoui16.pdf}, url = {https://proceedings.mlr.press/v49/elalaoui16.html}, abstract = {Given a weighted graph with N vertices, consider a real-valued regression problem in a semi-supervised setting, where one observes n labeled vertices, and the task is to label the remaining ones. We present a theoretical study of \ell_p-based Laplacian regularization under a d-dimensional geometric random graph model. We provide a variational characterization of the performance of this regularized learner as N grows to infinity while n stays constant; the associated optimality conditions lead to a partial differential equation that must be satisfied by the associated function estimate \widehatf. From this formulation we derive several predictions on the limiting behavior the function \fhat, including (a) a phase transition in its smoothness at the threshold p = d + 1; and (b) a tradeoff between smoothness and sensitivity to the underlying unlabeled data distribution P. Thus, over the range p ≤d, the function estimate \widehatf is degenerate and “spiky,” whereas for p≥d+1, the function estimate \fhat is smooth. We show that the effect of the underlying density vanishes monotonically with p, such that in the limit p = ∞, corresponding to the so-called Absolutely Minimal Lipschitz Extension, the estimate \widehatf is independent of the distribution P. Under the assumption of semi-supervised smoothness, ignoring P can lead to poor statistical performance; in particular, we construct a specific example for d=1 to demonstrate that p=2 has lower risk than p=∞due to the former penalty adapting to P and the latter ignoring it. We also provide simulations that verify the accuracy of our predictions for finite sample sizes. Together, these properties show that p = d+1 is an optimal choice, yielding a function estimate \fhat that is both smooth and non-degenerate, while remaining maximally sensitive to P. } }
Endnote
%0 Conference Paper %T Asymptotic behavior of \ell_p-based Laplacian regularization in semi-supervised learning %A Ahmed El Alaoui %A Xiang Cheng %A Aaditya Ramdas %A Martin J. Wainwright %A Michael I. Jordan %B 29th Annual Conference on Learning Theory %C Proceedings of Machine Learning Research %D 2016 %E Vitaly Feldman %E Alexander Rakhlin %E Ohad Shamir %F pmlr-v49-elalaoui16 %I PMLR %P 879--906 %U https://proceedings.mlr.press/v49/elalaoui16.html %V 49 %X Given a weighted graph with N vertices, consider a real-valued regression problem in a semi-supervised setting, where one observes n labeled vertices, and the task is to label the remaining ones. We present a theoretical study of \ell_p-based Laplacian regularization under a d-dimensional geometric random graph model. We provide a variational characterization of the performance of this regularized learner as N grows to infinity while n stays constant; the associated optimality conditions lead to a partial differential equation that must be satisfied by the associated function estimate \widehatf. From this formulation we derive several predictions on the limiting behavior the function \fhat, including (a) a phase transition in its smoothness at the threshold p = d + 1; and (b) a tradeoff between smoothness and sensitivity to the underlying unlabeled data distribution P. Thus, over the range p ≤d, the function estimate \widehatf is degenerate and “spiky,” whereas for p≥d+1, the function estimate \fhat is smooth. We show that the effect of the underlying density vanishes monotonically with p, such that in the limit p = ∞, corresponding to the so-called Absolutely Minimal Lipschitz Extension, the estimate \widehatf is independent of the distribution P. Under the assumption of semi-supervised smoothness, ignoring P can lead to poor statistical performance; in particular, we construct a specific example for d=1 to demonstrate that p=2 has lower risk than p=∞due to the former penalty adapting to P and the latter ignoring it. We also provide simulations that verify the accuracy of our predictions for finite sample sizes. Together, these properties show that p = d+1 is an optimal choice, yielding a function estimate \fhat that is both smooth and non-degenerate, while remaining maximally sensitive to P.
RIS
TY - CPAPER TI - Asymptotic behavior of \ell_p-based Laplacian regularization in semi-supervised learning AU - Ahmed El Alaoui AU - Xiang Cheng AU - Aaditya Ramdas AU - Martin J. Wainwright AU - Michael I. Jordan BT - 29th Annual Conference on Learning Theory DA - 2016/06/06 ED - Vitaly Feldman ED - Alexander Rakhlin ED - Ohad Shamir ID - pmlr-v49-elalaoui16 PB - PMLR DP - Proceedings of Machine Learning Research VL - 49 SP - 879 EP - 906 L1 - http://proceedings.mlr.press/v49/elalaoui16.pdf UR - https://proceedings.mlr.press/v49/elalaoui16.html AB - Given a weighted graph with N vertices, consider a real-valued regression problem in a semi-supervised setting, where one observes n labeled vertices, and the task is to label the remaining ones. We present a theoretical study of \ell_p-based Laplacian regularization under a d-dimensional geometric random graph model. We provide a variational characterization of the performance of this regularized learner as N grows to infinity while n stays constant; the associated optimality conditions lead to a partial differential equation that must be satisfied by the associated function estimate \widehatf. From this formulation we derive several predictions on the limiting behavior the function \fhat, including (a) a phase transition in its smoothness at the threshold p = d + 1; and (b) a tradeoff between smoothness and sensitivity to the underlying unlabeled data distribution P. Thus, over the range p ≤d, the function estimate \widehatf is degenerate and “spiky,” whereas for p≥d+1, the function estimate \fhat is smooth. We show that the effect of the underlying density vanishes monotonically with p, such that in the limit p = ∞, corresponding to the so-called Absolutely Minimal Lipschitz Extension, the estimate \widehatf is independent of the distribution P. Under the assumption of semi-supervised smoothness, ignoring P can lead to poor statistical performance; in particular, we construct a specific example for d=1 to demonstrate that p=2 has lower risk than p=∞due to the former penalty adapting to P and the latter ignoring it. We also provide simulations that verify the accuracy of our predictions for finite sample sizes. Together, these properties show that p = d+1 is an optimal choice, yielding a function estimate \fhat that is both smooth and non-degenerate, while remaining maximally sensitive to P. ER -
APA
El Alaoui, A., Cheng, X., Ramdas, A., Wainwright, M.J. & Jordan, M.I.. (2016). Asymptotic behavior of \ell_p-based Laplacian regularization in semi-supervised learning. 29th Annual Conference on Learning Theory, in Proceedings of Machine Learning Research 49:879-906 Available from https://proceedings.mlr.press/v49/elalaoui16.html.

Related Material